Wearing a CGM Without Diabetes: What the First Meta-Analysis Actually Found

What the 2026 Meta-Analysis Actually Pooled

Liao and colleagues searched PubMed, the Cochrane Library, Web of Science, and Embase for studies published through December 2025. Two reviewers independently screened titles and abstracts, assessed full texts, and extracted data. They included 23 studies: 7 randomized controlled trials, 1 cohort study, 7 prospective observational studies, 4 cross-sectional studies, and 4 clinical trials, covering 1,074 non-diabetic participants across 11 countries.

Two things about that composition matter before you read any result. First, this is all human data. There is no preclinical or animal evidence mixed in, which is a genuine strength relative to a lot of the wellness-adjacent glucose literature. Second, only 7 of the 23 studies were randomized controlled trials, and the remaining 16 were observational or single-arm designs. A pooled estimate built substantially on observational work carries different weight than one built on a stack of large randomized trials.

The included studies clustered around lifestyle aims rather than diagnosis. The review reports overlapping objectives across the 23 studies, using its own category labels: glycemic optimization in 9 of 23 (39%), weight management in 9 of 23 (39%), dietary behavior modification in 12 of 23 (52%), and adherence enhancement in 5 of 23 (22%). In other words, most of this literature asked whether a sensor changes what a person does, not whether a sensor reveals something a blood draw would have missed.

The Results: One Significant Pooled Outcome, and Who It Applied To

Here is the outcome picture as the review reports it.

|---|---|---|

### The glucose result, and the subgroup split inside it

The glucose finding is real and worth taking seriously. An SMD of -0.54 is a moderate effect size, and an I2 of 0% means the included studies pointed in a consistent direction rather than being averaged out of a mess of conflicting findings. But look at the confidence interval: -1.02 to -0.07. The upper bound sits very close to zero. That is a result clearing the significance threshold without much room to spare, which is normal for a pooled analysis of this size and a reason to describe it as promising rather than settled.

More important is who the benefit showed up in. The authors state that CGM improved glycemic control in individuals with prediabetes, whereas no appreciable glycemic benefit was observed in healthy normoglycemic populations. That single sentence reframes the entire consumer pitch. The headline pooled number is not evidence that a metabolically healthy person will see their glucose move by wearing a sensor. In this evidence base, it did not.

### The null results deserve equal billing

Body mass index did not differ significantly. A sensor that nudged glucose readings did not, in this pooled evidence, translate into a measurable body-composition change. The authors go further in their conclusion, writing that CGM is unlikely to achieve weight management goals independently without accompanying behavioral interventions.

Measurement accuracy showed no significant difference between groups, with a mean difference of -3.90 mg/dL and a P value of 0.40. Note what that comparison is and is not. A non-significant result is not a demonstration that a consumer sensor reads the same as a venous draw. The authors themselves say evidence on measurement accuracy in this population remains limited.

Glycemic variability was handled by qualitative synthesis rather than pooling. The review examined variability metrics including mean amplitude of glycemic excursion, standard deviation, and continuous overlapping net glycemic action over one hour, and concluded the impact of CGM on those metrics was context-dependent. Anyone citing a clean pooled variability result from this paper is citing something that is not in it.

### The confounder running through all of it

The least convenient limitation for anyone selling a sensor is structural. Most of the included studies targeted health promotion through lifestyle interventions, which means the sensor came wrapped in coaching, dietary guidance, or a structured protocol. If glucose improved, the sensor may have caused it, the program may have caused it, or the two may only work together. The authors' own framing is the most defensible read available: CGM should be positioned as a precision biofeedback tool integrated within structured lifestyle programs. Biofeedback, not diagnostics. Integrated, not standalone.

Other constraints are worth carrying with you. The review notes that sensitivity analysis was not performed because of the limited number of included studies. Definitions of what counts as non-diabetic varied across studies. And the authors characterize the dietary-behavior evidence as short-term, which means nothing here speaks to what happens after a year of wear, or to whether any glucose improvement persists once the sensor comes off.

### The adherence finding, and why it is the most practical one

Buried under the effect sizes is a number with more day-to-day relevance than any of them. The review reports consistently high sensor wear rates above 90% across multiple studies, and found CGM use associated with higher behavioral adherence and specific dietary modifications.

That is unusual. Most self-tracking interventions leak participants badly. Food logs, symptom diaries, and manual weigh-ins all tend to decay within weeks. A device people keep wearing at above 90% is doing something that questionnaires and journals do not, and for anyone building a tracking habit, adherence is the constraint that actually binds. Data you never collect cannot help you or your clinician.

The uncomfortable corollary is that high adherence to a device is not the same as high adherence to a useful record. Ninety-six datapoints a day of raw glucose, with no note about what you ate, how you slept, or whether you trained that morning, is a large volume of uninterpretable numbers. The adherence finding is an argument for building a logging habit around the sensor, not an argument that the sensor alone is enough.

What the ADA Standards of Care 2026 Did and Did Not Say

The ADA Standards of Care in Diabetes 2026, published 8 December 2025 in Diabetes Care volume 49, supplement 1, pages S150 to S165, addresses consumer sensors directly in its diabetes technology chapter. It states plainly that "anyone can purchase the OTC-CGM devices, including those without diabetes or with prediabetes who wish to assess their glycemic responses to their lifestyles, including the effects of food choices and exercise."

That is an acknowledgment of a market reality, not an endorsement. The same edition contains no recommendation for CGM as a screening or diagnostic tool for prediabetes or diabetes. Every CGM recommendation in the chapter applies to people who already carry a diabetes diagnosis. The chapter now distinguishes three device categories: real-time CGM, over-the-counter CGM, and professional CGM.

The 2026 edition did broaden CGM within that diagnosed population. Recommendation 7.15 recommends CGM at diabetes onset and anytime thereafter for children, adolescents, and adults with diabetes who are on insulin therapy (evidence grade A), on noninsulin therapies that can cause hypoglycemia (grade C), and on any diabetes treatment where CGM helps in management (grade C), with device selection based on the individual's circumstances, preferences, and needs (grade E). The extension past insulin users is the news. The absence of any screening recommendation for people without diabetes is, for this article's purposes, the bigger news.

The FDA's June 2026 Clearance, Read Carefully

On 12 June 2026, the FDA cleared Dexcom's Stelo Glucose Biosensor System over the counter for people two years of age and older who do not use insulin, the first over-the-counter CGM cleared for children. Stelo had previously been cleared over the counter for adults 18 and older in March 2024. Each sensor lasts up to 15 days before it must be replaced, though wear time may be shorter in pediatric users than in adults, and the paired app displays glucose values and trends every 15 minutes, roughly 96 datapoints per day.

Two points in that paragraph carry weight. Cleared is a regulatory finding about a device, not a clinical recommendation about who should wear one. And the clearance itself carries the caveat that the device is not intended for users to take medical action on readings without consulting a health care professional, including any change to a medication regimen.

Reading the three 2026 items together, they are not in conflict. A device can be legally available over the counter, produce a modest measurable effect on mean glucose inside a structured program and mainly in people with prediabetes, and still have no guideline body recommending it as a screening test. All three are true at once.

The 140 mg/dL Question, and How to Answer It Calmly

The most common scenario a coach or practitioner will face is a client who eats a bowl of rice, sees a post-meal number in the 140s, screenshots it, and asks whether they are prediabetic.

The sourced answer is that no current guideline supports drawing that conclusion from an over-the-counter sensor. The ADA Standards of Care 2026 issues no recommendation for CGM as a screening or diagnostic tool. The 2026 meta-analysis found no significant between-group difference in measurement accuracy and calls that accuracy evidence limited. A single interstitial reading, from a device with a known lag behind blood glucose, without a validated cutoff for non-diabetic self-screening, is not something to reason backward from.

What the reading is good for is context. It is one datapoint in a behavioral feedback loop, and it becomes informative only when it is logged next to what was eaten, when it was eaten, how the person slept, and whether they trained. That is the platform's argument in a sentence: the value is in the annotated series, not the raw number. Whether any pattern in that series means anything clinically is a determination for a licensed clinician working from a proper blood draw and the person's full history. Our existing coverage of [wearables paired with bloodwork](/blog/wearable-bloodwork-peptide-tracking-2026) already framed CGM as continuous context rather than a replacement for labs, and the 2026 evidence supports that framing.

What to Track and How to Log It

If you wear a sensor, the markers below are the ones commonly discussed alongside it in the metabolic literature. This is not a recommendation to test anything. It is a list of what people track so you can organize your own record and review it with your provider.

Alongside the labs, the sensor context worth logging is behavioral: meal composition and timing, sleep duration and quality, training type and time of day, alcohol, and illness. Those are the variables that make a glucose curve interpretable. If your care involves GLP-1 medications, the same principle applies and the relevant panel is covered in our [GLP-1 lab monitoring guide](/blog/glp1-lab-monitoring-panel). You can browse plain-language explanations of each test in the full [biomarker library](/biomarkers).

Ninety-six readings a day is a lot of noise and not much signal on its own. The approach that matches what the evidence supports is to treat the sensor as biofeedback inside a structure, which is precisely how the meta-analysis authors positioned it. That means logging context with every reading you care about rather than staring at the live graph, keeping your fasting blood draws consistent so the venous reference stays comparable over time (our [fasting blood work guide](/blog/fasting-blood-work-guide) covers the conditions that matter), reviewing patterns across weeks rather than reacting to single spikes, and bringing the annotated record to your clinician instead of a screenshot. Nothing in the 2026 evidence supports self-diagnosing from sensor output, and nothing in it supports changing a medication or protocol based on what an app shows.

The reason a structured record helps is friction. Storing sensor context, lab values, and protocol notes in one place, and being able to see them on a shared timeline, is the difference between a habit that survives past week three and a camera roll full of screenshots. MyProtocolStack organizes and visualizes that data. It does not interpret it, and it is not a substitute for your clinician's judgment.

[Log your CGM context alongside your lab trends in one record with MyProtocolStack.](/auth/login?mode=signup)

Frequently Asked Questions

Does wearing a CGM help if you do not have diabetes?

The first systematic review and meta-analysis focused on this population (Liao et al., European Journal of Medical Research 2026;31:397) pooled 23 studies and 1,074 participants and found that CGM use significantly improved mean blood glucose versus controls, SMD -0.54 (95% CI -1.02 to -0.07, P = 0.03), with low heterogeneity (I2 = 0%). The critical qualifier is that the authors report improved glycemic control in individuals with prediabetes but no appreciable glycemic benefit in healthy normoglycemic populations. The review found no significant difference in body mass index or measurement accuracy, describes the accuracy and glycemic variability evidence as limited, and positions CGM as a precision biofeedback tool integrated within structured lifestyle programs rather than a standalone intervention.

Can a CGM tell me if I have prediabetes?

No current guideline supports that use. The ADA Standards of Care in Diabetes 2026, published 8 December 2025, acknowledges that anyone can purchase over-the-counter CGM devices including people without diabetes or with prediabetes, but it issues no recommendation for CGM as a screening or diagnostic tool for prediabetes or diabetes. Every CGM recommendation in that chapter applies to people who already have a diabetes diagnosis. Screening and diagnosis are clinical determinations your provider makes using validated testing and your full history, not something to infer from a consumer sensor reading.

What should I make of a 140 mg/dL reading after a meal?

Treat it as one datapoint in a behavioral feedback loop, not as evidence about your metabolic status. No guideline body currently supports diagnosing anything from an over-the-counter sensor, and the 2026 meta-analysis found no significant between-group difference in measurement accuracy (MD -3.90 mg/dL, P = 0.40) while calling that accuracy evidence limited. The reading becomes informative only when it is logged with meal composition, timing, sleep, and training context and reviewed over weeks. Any interpretation of what a pattern means belongs with a licensed clinician working from a proper blood draw.

Did the meta-analysis show CGM causes weight loss in people without diabetes?

No. Pooled body mass index showed no significant difference (SMD -0.25, 95% CI -0.63 to 0.12, P = 0.19). The authors conclude that while CGM demonstrates high feasibility in modifying short-term dietary behaviors, it is unlikely to achieve weight management goals independently without accompanying behavioral interventions. Because most included studies delivered the sensor alongside coaching or a structured lifestyle program, the review cannot cleanly separate the device's contribution from the program's.

What did the FDA actually clear in June 2026?

On 12 June 2026 the FDA cleared Dexcom's Stelo Glucose Biosensor System over the counter for people two years of age and older who do not use insulin, making it the first over-the-counter CGM cleared for children. Stelo had previously been cleared over the counter for adults 18 and older in March 2024. Each sensor lasts up to 15 days, with wear time potentially shorter in pediatric users than adults, and the app displays glucose values and trends every 15 minutes. The clearance carries the caveat that users should not take medical action on readings, including medication changes, without consulting a health care professional.

What can a CGM not tell me that bloodwork can?

A sensor measures interstitial glucose and nothing else. It cannot measure fasting insulin, or C-peptide, or triglycerides, or HbA1c, which reflects roughly 90 days of average glucose. It also cannot produce calculated markers like HOMA-IR and the TyG index that depend on insulin and lipid inputs. Continuous glucose and periodic bloodwork answer different questions, which is why the most useful record pairs sensor context with consistently drawn fasting labs rather than treating either one as a substitute for the other.

Sources

1. Liao X, Li Y, Tang S, et al. "Continuous glucose monitoring in non-diabetic populations: a systematic review of observational and interventional studies with meta-analysis." European Journal of Medical Research 2026;31:397, published 26 January 2026. PMID 41588451. Retrieved via PubMed. https://doi.org/10.1186/s40001-026-03920-0

2. American Diabetes Association, "7. Diabetes Technology: Standards of Care in Diabetes 2026," Diabetes Care 2026;49(Suppl 1):S150-S165, published 8 December 2025, DOI 10.2337/dc26-S007. https://pmc.ncbi.nlm.nih.gov/articles/PMC12690173/

3. American Diabetes Association, "7. Diabetes Technology: Standards of Care in Diabetes 2026," journal landing page. https://diabetesjournals.org/care/article/49/Supplement_1/S150/163922/7-Diabetes-Technology-Standards-of-Care-in

4. U.S. Food and Drug Administration, "FDA Clears First Over-the-Counter Continuous Glucose Monitor for Children," 12 June 2026. https://www.fda.gov/news-events/press-announcements/fda-clears-first-over-counter-continuous-glucose-monitor-children

5. "The efficacy of using continuous glucose monitoring as a behaviour change tool in populations with and without diabetes: a systematic review and meta-analysis of randomised controlled trials." International Journal of Behavioral Nutrition and Physical Activity 2024. Retrieved via PubMed Central. https://pmc.ncbi.nlm.nih.gov/articles/PMC11668089/

*MyProtocolStack is a tracking and education tool, not medical advice, diagnosis, or treatment, and you should always consult a qualified healthcare professional before making any changes to your health protocol.*